Applying parametric models for survival analysis of gastric cancer

Cox Proportional Hazard model is the most popular tool to analyze the effects of covariates on survival time, although under certain circumstances parametric models may offer advantages over Cox model. In this study using Cox regression and alternative parametric models (such as Weibull, Exponential and Lognormal models) the prognostic factors affecting the survival of patients with gastric cancer were evaluated and comparisons for finding the best model were made.

To determine the independent prognostic factors reducing survival time for gastric cancer, we compared the parametric and semi-parametric methods applied to patients registered in one cancer registry center in the southern part of Iran. Using Akaike Information Criterion, we compared parametric and Semi-parametric models.

Two-hundred sixty six (60.2 %) out of 442 patients died during the period. Data analysis results using Cox and Parametric models are nearly similar. According to Akaike Information Criterion, Weibull (AIC=848) and Exponential (AIC=850) models had the best fitness for survival analysis. Patients (ages 60-75 and above) at time of diagnosis had an increased risk of death followed by those with poor differentiated grade and the distant metastasis (P<0.05).

Although the Hazard Ratio in Cox and parametric models are not completely similar, according to Akaike Information Criterion, Weibull and Exponential models are the most favorable ones for survival analysis.